Study Of Denoising And Segmentation Algorithms For SAR Image

Synthetic aperture radar (SAR) is an important use for military reconnaissance and civil activity, so it has practical meaning and application prospect to study feature extraction and object recognition method for SAR images. In this paper, we study the preprocesses which are the indispensable steps in the process of SAR including the RCS reconstruction and the segmentation of SAR in theory, and validate them by experimental results.We study the method of SAR image filtering and two methods are presented. First, a novel approach based on the correlated neighborhood model is proposed which describes the local property of the image. Second, a new wavelet-based denoising method without free parameters is proposed which is based on the sparseness and decorrelation properties of the discrete wavelet transform. Finally, we make a comparison of the characteristics of the traditional filter based on the statistics of SAR image with those of the filter based on the wavelet transform in the process of SAR image denoising. Two segmentation methods of SAR image are proposed. First, the genetic algorithm is used in the SAR image segmentation. Second, we proposed a SAR image segmentation based on the criterion of likelihood difference. Simulations verify the effectiveness of this method.A novel pixel-level image fusion scheme was presented based on the advantages of the wavelet transform. The wavelet transform was used to perform a multiscale decomposition of each image. The multiple operators and rules are used in the different frequence regions on the different scales. The experimental results show that the fusion scheme is effective for different image data.The paper is supported by "863" High-Tech program and National Natural Science Foundation of China.